面向业务流程的服务划分，推荐及管理关键技术研究

SOA is a very important mechanism to support flexible business process. In this research, inspired by the relationship between service and business process, comprehensive analysis is conducted from viewpoint of service and its evolution with regard to the business process. Particularly three objectives are identified, i.e., service identification, service recommendation and service governance. First the association rule is employed to describe the inter-dependency among services, based on which a semantic analysis is utilized to describe the dependency among services. Afterwards service identification is realized by using clustering methods. After service is identified, the next important task is to properly recommend it to the users. In this research, the recommendation process uses the implicit selection ratio of services on particular process as a preference hints. And then a collaborative filtering based re-raking mechanism is designed to support appropriate service recommendation. Furthermore, since business is always changing from time to time, it becomes necessary for the service to evolve gradually accordingly. In this research, a service lifecycle management framework is discussed to support service evolution to make sure the service is always consistent with the requirement of business process. As such the main research objectives are as follow: 1) business process aware service identification; 2) business constraint oriented service recommendation; 3) stakeholder based service evolution and management. Finally a prototype of for service governance supporting service identification and recommendation will be conducted and case study and field study will be included for validating the proposed mechanism.

本课题依托项目组之前在服务计算领域的研究成果，探索从业务流程的角度去研究服务及演化过程，围绕服务与之所处业务流程的关系，开展服务的划分、推荐和治理的研究。通过研究服务组合的历史并采用关联规则挖掘分析服务依存关系，并在此基础上依靠层次分析法逐步抽取服务候选单元。为了将合理划分的服务更好地参与到业务流程中，在关联规则分析的基础上，提出了隐式服务偏好评分方法，利用改进的协同过滤方法实现服务的推荐。同时，进一步研究服务随着业务改变而不断演化的生命周期管理机制，引入可量化的利益相关人分析框架，使得服务始终能够与所涉及的业务流程保持关联性，确保服务始终处与一个合理的粒度。具体研究内容包括：1）基于业务流程的服务划分技术；2）面向业务规则的服务推荐技术；3）基于利益相关者的服务治理技术，并最终搭建一个面向业务流程的服务综合管理原型系统，验证方法和技术的有效性。

基于服务的组件技术使得业务流程能够快速的进行搭建，从而提高企业在市场上的竞争优势，如何高效的对服务进行组织是目前服务计算领域最重要的一个挑战之一。对服务进行治理首先需要获得一个大小和粒度都合适的服务。服务一般会用于业务流程，因此，如何从业务流程角度识别服务是一个非常重要的挑战。本课题针对这个情况，提出了一种基于业务流程的需求依赖建模方法，并基于此设计了一个候选服务的划分和选择框架。在获得服务之后，如何将服务进行推荐也是一个必不可少的工作，因此，本课题将现有推荐技术与深度学习技术相结合，提出了一个基于神经网络的上下文相关推荐方法，并在用于服务推荐时，还重点考虑服务的可信度和防攻击能力。最后，对于已存在的网络在线服务，本课题提出了一个基于扩展TAM模型的关键影响因子建模方法，分析网络服务被用户接受和使用的原因和趋势。